Inventive Process as a Recombinant Search over Complex Landscape: Evidence from the Disk Drive Industry

The inventive process has been often modeled as a bounded iterative trial-and-error recombinant search over complex landscape. Our main research question is whether such approximation is empirically valid. To investigate it, we develop a single-industry measure of invention interdependence and provide a relatively direct test of the canonical NK model. Our findings indicate that the NK model correctly predicts most of the empirical patterns. We also find that the consistency between the empirical estimates and the model predictions deteriorates with expanded definition of industry boundaries. Our results suggest that models representing iterative trial-and-error recombinant search are applicable as approximations of the inventive process when one looks at single mature industries where most of the knowledge originates from within the same technological domain. When inventors draw from a broad knowledge base and fundamentally new knowledge is created then applicability of simple models of recombinant search may be limited.